Data Science at Pluralsight: Our Principles
Introducing our Principles
Our Data Science and Machine Learning practice at Pluralsight has grown quickly but what we value and how we work has been built in from the beginning. Our work influences decisions, both big and small, and our algorithms impact how our customers experience our product everyday. We want to create and maintain a discipline we are proud of, one where we respect each other and our differences, and one that makes us happy and excited to come to work each day. But, how can we be conscientious about scaling our Principles with the growth of our discipline during a time of growth, rapid hiring, and practitioners becoming spread throughout the business?
One of the most important ways to maintain a set of values is to talk about them and you can’t talk about them until you know them. The best way to get clear on what they are is to write them down. We believe that the best way to define what we value is to do so in terms of observable behaviors. Our engineers established their values early on as well, and we have adopted a similar format and approach as they have. If you’re curious, check out Pluralsight Engineering: How We Work.
In this post we’ll share how we established our values for Data Science at Pluralsight and what they are. In future posts, we’ll go deeper on some of these values, sharing how we apply them in practice.
How We Established our Principles
When we began our initial effort to codify our values, we thought about how we would do so and concluded that there were three essential elements needed in the process:
- We should be clear on why we are doing this
- We must co-create our Principles with everyone in our discipline
- We expect the process to be iterative and the principles to be evolving.
What do our Principles do for us?
When a team is small, you can often get away with hiring people who think and work in ways that are aligned with what the current team values. While it continues to be important to look to our those values as we evaluate candidates in the hiring process, we knew that scaling and maintaining them would take more than that. Overall, we wanted a set of principles that do the following:
They reflect what we value
While our data projects change, there is a certain craftsmanship mindset that permeates all of our work around transparency, reproducibility, and quickly eliciting feedback. The principles allow us to codify what we value to not only remind ourselves but to make it clear to the community and future team members.
They serve as a compass in our daily work
Working with cross-functional teams in a fast growing company, it’s very easy to be pulled into different directions and lose focus on how to add the most value. The principles empower us to step back, use them as a reference to reevaluate our approach and strategy, examine the quality of our work, and make the right decisions.
They enable us to scale with consistency
A good set of principles, and adherence to those principles, is key to having a craft scale with consistency. They both help create high-level uniformity and allow for experimentation (around tooling and process) by individual practitioners, while also maintaining a group identity and bounds which is useful for managing and growing deliberately.
They show us what to aspire to
They provide a guiding light and help build the craft of data professionals. Data practitioners should feel deep pride in their work and their field, and established principles (whether within a company or an industry) help provide a source of camaraderie and pride in committing to something bigger, not sacrificing what we hold dear, and helping us strive toward professional excellence.
Our Principles were Co-created
We believe that our Principles should truly reflect our practitioners’ needs and aspirations in conducting Data Science and Machine Learning work. We knew that they needed to be developed ground-up from practitioners rather than foisted down from leaders. Co-creating the principles together would create the highest likelihood that everyone would embrace them.
We also recognized that our discipline comprises a breadth of skills and roles ranging from advanced analytics to machine learning engineering, each with differing day to day tasks and responsibilities. It is important that the principles reflect the voices and collective learnings of all of these team members.
That said, we also knew that we couldn’t start with a blank sheet of paper. So we started with a small group of people and a large stack of sticky notes. This got us to some core themes and a straw-man document to work with.
We then engaged all of our Data Science practitioners and conducted a substantial workshop in which we broke out into small groups and picked apart and built back up the document. We brought together all of the suggested additions, edits, input on what did and didn’t resonate and ended with another draft.
Over the following months, we continued to refine, reword, and revisit. It took us almost six months of iteration to get to the documented Principles that we now reference.
Through the process of co-creating our Principles with everyone in our discipline, ideas were shared and improved together and everyone’s voice was heard and challenged. At the end, the principles were a collection of most valued guidelines that resonate with each practitioner.
Our Principles are Foundational but Evolving
As noted, it took almost six months to produce a codified set of principles that resonated across our discipline. While we expect the majority of our principles to be foundational and remain largely unchanged, we also recognize that our discipline is rapidly changing. Not only that, but because the company is growing we must adapt to changes in our organization and evolve how we work.
Certainly in a year’s time, the org may be a level up in terms of their Data Science practices and we’ll have to evolve the principles to further elevate the craft at Pluralsight, reflective not only of the enhanced ability for Data Science collectively, but also reflective of the needs of the customer and business.
— Levi Thatcher, Principal Data Scientist
Because we value growth and learning, we expect to consciously revisit the principles frequently and ask practitioners for feedback on what principles were followed, referred to, or ignored in their day to day work.
As everyone and every team have to constantly learn and grow, we need to ensure that the principles stay relevant to our practices. I’d treat them as our data science models, which need to be fine-tuned or re-trained over time, to adapt themselves to their targeted population.
— Shan Huang, Principal Data Scientist
Even though the principles may change, we feel it’s important to put a stake in the ground now. How many times have you heard or read that you need to write down your goals to achieve them? The same holds true for values — whether for yourself, your team, or your company. Having gone through the work to put ours to paper, we have seen both an upgrade in the work of our discipline and an increase in community commitment to them.
What are our Principles?
So, after all the work to get here, what ARE our Principles? What they look like for us is a living document. We have organized around three core elements that relate to the different aspects of our work. These are:
- What we work on. These values guide us to align our work with the business strategy, ensure we have an understanding of the outcomes we are driving, and how the solutions we build address the needs of our customers.
- How we work. Through experience, we have seen patterns of what has created successful outcomes and what has caused us to stumble. This set of values guide us to embrace scientific rigor, start with simple solutions, deliver iteratively, and communicate effectively.
- What is important to us. We collectively hold ourselves and our work to a high standard and we look after that collectively. We value collaboration, a culture of continuous learning, and growth over comfort. These values articulate this.
Within each of these core elements, are our Principles. We believe that principles and values are only useful if we can describe what they look like. To that end, associated with each Principle are a set of behaviors that articulate what we do, what we encourage, and what we avoid.
If you haven’t, check out the details of our Principles here and look for future posts that demonstrate how these values manifest themselves in our work. This is an evolving document and, as lifelong learners, we welcome comments and feedback.
